Certain groups of people (defined by ethnicity, SES, sexual orientation, etc) have societal pressures placed upon them such that they are at a disadvantage compared to other groups. This means that their baseline level of performance in certain areas is lower than the non-disadvantaged groups due to factors that are largely outside of their control, so even if the change in performance over time is equal to or greater than the non-disadvantaged groups, the final result may be less.
In a certainly less-than-perfect MCAT-based analogy, a lesbian black female applicant who lives with her single mother in low income city housing with her two brothers, due to the societal pressures of being LGBT, black, female, economically disadvantaged, and growing up in a single-parent household, may be "starting" college with an MCAT score of 16. However, she works very hard, studies a lot, and by the time she finally takes her MCAT, she ends up with a 27. That is an 11 point gain.
In contrast, let's take white, upper middle class male student whose parents are both Ivy League graduates (one is a cardiologist, the other a patent attorney) and has one older brother, both of whom attended Exeter. This applicant, because of the advantages afforded to him by having a wealthy and stable family and attending a top-notch private high school, as well as having been enrolled in private schools since preschool, "starts" college with an MCAT score of 25. He goes to college and does well. When he finally takes the MCAT, he manages a 35. That is a 10 point gain.
We see that even though the second applicant has a higher final score, the first applicant has a greater "increase" in score over the same period of time. These scores do not exist in a vacuum, and medical schools know it. They are ultimately looking for aptitude and potential, and having someone who has a high point gain, particularly when facing heavy adversity, is very valuable to them, because that potential can be utilized to generate a great physician. This doesn't mean that either applicant is "better" than the other (the second applicant has a great score and will likely get into a medical school), just that the starting points are different (again, initial conditions are not created by the applicants in either situation), and thus the end points, given similar rates of change, are likely to be different as well, so that has to be taken into account.